2010-04-25 07:03:10

by Xiao Guangrong

Subject: [PATCH v2 1/10] KVM MMU: fix for calculating gpa in invlpg code

If the guest is 32-bit, we should use 'quadrant' to adjust gpa
offset

Changlog v2:
- when level is PT_DIRECTORY_LEVEL, the 'offset' should be
'role.quadrant << 8', thanks Avi for point it out

Signed-off-by: Xiao Guangrong <[email protected]>
---
arch/x86/kvm/paging_tmpl.h | 13 +++++++++++--
1 files changed, 11 insertions(+), 2 deletions(-)

diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
index d0cc07e..83cc72f 100644
--- a/arch/x86/kvm/paging_tmpl.h
+++ b/arch/x86/kvm/paging_tmpl.h
@@ -478,9 +478,18 @@ static void FNAME(invlpg)(struct kvm_vcpu *vcpu, gva_t gva)
((level == PT_DIRECTORY_LEVEL && is_large_pte(*sptep))) ||
((level == PT_PDPE_LEVEL && is_large_pte(*sptep)))) {
struct kvm_mmu_page *sp = page_header(__pa(sptep));
-
+ int offset = 0;
+
+ if (PTTYPE == 32) {
+ if (level == PT_DIRECTORY_LEVEL)
+ offset = PAGE_SHIFT - 4;
+ else
+ offset = PT64_LEVEL_BITS;
+ offset = sp->role.quadrant << offset;
+ }
pte_gpa = (sp->gfn << PAGE_SHIFT);
- pte_gpa += (sptep - sp->spt) * sizeof(pt_element_t);
+ pte_gpa += (sptep - sp->spt + offset) *
+ sizeof(pt_element_t);

if (is_shadow_present_pte(*sptep)) {
rmap_remove(vcpu->kvm, sptep);
--
1.6.1.2

2010-04-25 09:46:37

by Avi Kivity

Subject: Re: [PATCH v2 1/10] KVM MMU: fix for calculating gpa in invlpg code

On 04/25/2010 10:00 AM, Xiao Guangrong wrote:
> If the guest is 32-bit, we should use 'quadrant' to adjust gpa
> offset
>
> Changlog v2:
> - when level is PT_DIRECTORY_LEVEL, the 'offset' should be
> 'role.quadrant<< 8', thanks Avi for point it out
>
> Signed-off-by: Xiao Guangrong<[email protected]>
> ---
> arch/x86/kvm/paging_tmpl.h | 13 +++++++++++--
> 1 files changed, 11 insertions(+), 2 deletions(-)
>
> diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
> index d0cc07e..83cc72f 100644
> --- a/arch/x86/kvm/paging_tmpl.h
> +++ b/arch/x86/kvm/paging_tmpl.h
> @@ -478,9 +478,18 @@ static void FNAME(invlpg)(struct kvm_vcpu *vcpu, gva_t gva)
> ((level == PT_DIRECTORY_LEVEL&& is_large_pte(*sptep))) ||
> ((level == PT_PDPE_LEVEL&& is_large_pte(*sptep)))) {
> struct kvm_mmu_page *sp = page_header(__pa(sptep));
> -
> + int offset = 0;
> +
> + if (PTTYPE == 32) {
> + if (level == PT_DIRECTORY_LEVEL)
> + offset = PAGE_SHIFT - 4;
> + else
> + offset = PT64_LEVEL_BITS;
> + offset = sp->role.quadrant<< offset;
> + }
>

The calculation is really

shift = (PT32_LEVEL_BITS - PT64_LEVEL_BITS) * level;

(and please don't use a variable called offset to hold a shift count)


--
error compiling committee.c: too many arguments to function

2010-04-26 03:13:25

by Xiao Guangrong

Subject: Re: [PATCH v2 1/10] KVM MMU: fix for calculating gpa in invlpg code

Avi Kivity wrote:
> On 04/25/2010 10:00 AM, Xiao Guangrong wrote:
>> If the guest is 32-bit, we should use 'quadrant' to adjust gpa
>> offset
>>
>> Changlog v2:
>> - when level is PT_DIRECTORY_LEVEL, the 'offset' should be
>> 'role.quadrant<< 8', thanks Avi for point it out
>>
>> Signed-off-by: Xiao Guangrong<[email protected]>
>> ---
>> arch/x86/kvm/paging_tmpl.h | 13 +++++++++++--
>> 1 files changed, 11 insertions(+), 2 deletions(-)
>>
>> diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
>> index d0cc07e..83cc72f 100644
>> --- a/arch/x86/kvm/paging_tmpl.h
>> +++ b/arch/x86/kvm/paging_tmpl.h
>> @@ -478,9 +478,18 @@ static void FNAME(invlpg)(struct kvm_vcpu *vcpu,
>> gva_t gva)
>> ((level == PT_DIRECTORY_LEVEL&& is_large_pte(*sptep))) ||
>> ((level == PT_PDPE_LEVEL&& is_large_pte(*sptep)))) {
>> struct kvm_mmu_page *sp = page_header(__pa(sptep));
>> -
>> + int offset = 0;
>> +
>> + if (PTTYPE == 32) {
>> + if (level == PT_DIRECTORY_LEVEL)
>> + offset = PAGE_SHIFT - 4;
>> + else
>> + offset = PT64_LEVEL_BITS;
>> + offset = sp->role.quadrant<< offset;
>> + }
>>
>
> The calculation is really
>
> shift = (PT32_LEVEL_BITS - PT64_LEVEL_BITS) * level;
>

So, the offset is q << (PAGE_SHIFT - (PT32_LEVEL_BITS - PT64_LEVEL_BITS) * level - 2)?

As my origin thinking, for level = 2, one gfn has 4 shadow pages, so
every shadow page occupies PAGE_SIZE / 4, and sizeof(pt_element_t) is 4
in 32-bit guest, so the offset is:
q * (PAGE_SIZE/4/4) = q << (PAGE_SHIFT - 2 -2)
Obviously, my code is really ugly than your :-)

2010-04-26 07:39:32

by Avi Kivity

Subject: Re: [PATCH v2 1/10] KVM MMU: fix for calculating gpa in invlpg code

On 04/26/2010 06:10 AM, Xiao Guangrong wrote:
>
> Avi Kivity wrote:
>
>> On 04/25/2010 10:00 AM, Xiao Guangrong wrote:
>>
>>> If the guest is 32-bit, we should use 'quadrant' to adjust gpa
>>> offset
>>>
>>> Changlog v2:
>>> - when level is PT_DIRECTORY_LEVEL, the 'offset' should be
>>> 'role.quadrant<< 8', thanks Avi for point it out
>>>
>>> Signed-off-by: Xiao Guangrong<[email protected]>
>>> ---
>>> arch/x86/kvm/paging_tmpl.h | 13 +++++++++++--
>>> 1 files changed, 11 insertions(+), 2 deletions(-)
>>>
>>> diff --git a/arch/x86/kvm/paging_tmpl.h b/arch/x86/kvm/paging_tmpl.h
>>> index d0cc07e..83cc72f 100644
>>> --- a/arch/x86/kvm/paging_tmpl.h
>>> +++ b/arch/x86/kvm/paging_tmpl.h
>>> @@ -478,9 +478,18 @@ static void FNAME(invlpg)(struct kvm_vcpu *vcpu,
>>> gva_t gva)
>>> ((level == PT_DIRECTORY_LEVEL&& is_large_pte(*sptep))) ||
>>> ((level == PT_PDPE_LEVEL&& is_large_pte(*sptep)))) {
>>> struct kvm_mmu_page *sp = page_header(__pa(sptep));
>>> -
>>> + int offset = 0;
>>> +
>>> + if (PTTYPE == 32) {
>>> + if (level == PT_DIRECTORY_LEVEL)
>>> + offset = PAGE_SHIFT - 4;
>>> + else
>>> + offset = PT64_LEVEL_BITS;
>>> + offset = sp->role.quadrant<< offset;
>>> + }
>>>
>>>
>> The calculation is really
>>
>> shift = (PT32_LEVEL_BITS - PT64_LEVEL_BITS) * level;
>>
>>
> So, the offset is q<< (PAGE_SHIFT - (PT32_LEVEL_BITS - PT64_LEVEL_BITS) * level - 2)?
>

Ugh, what I meant was

shift = PAGE_SHIFT - (PT_LEVEL_BITS - PT64_LEVEL_BITS) * level;
offset = q << shift

so,

pte_gpa = (sp->gfn << PAGE_SHIFT) + offset + (spte - sp->spt) *
sizeof(pt_element_t)

No magic numbers please. Note it could work without the 'if (PTTYPE ==
32)'.


--
Do not meddle in the internals of kernels, for they are subtle and quick to panic.